Abstract: The aim of the paper is to find links between the dynamics of changes of statistical parameters and changes in spectral properties of the signal envelope of backscattered RF signals during the thermal process. We have shown previously that by using wavelet approximations these tendencies are better recognized in the case of the heating of a phantom sample than in the parallel analysis performed for a full signal envelope. Here we are currently expanding this statement to the case of heating a soft tissue sample in vitro. The shape parameter of the K- distributed random variable is considered as a statistical marker of temperature level changes. Additionally, the spectral properties of different levels of wavelet approximations are calculated and their sensitivity to temperature increase and decrease is demonstrated. Both approaches registering changes in temperature, are used in the case of the pork loin tissue sample in vitro, heated by an ultrasound beam with a different power.